31 results about "Nonlinear classification" patented technology

The invention relates to the technical field of intrusion detection, in particular to a network threat detection method and system, and the method comprises the following steps: building a threat path logic diagram through collecting field dependency items, action trigger timestamp items and action propagation hop count items of an attack behavior chain, and matching field dependency items among nodes based on a graph theory algorithm to obtain a threat path logic diagram; and detecting a mutual exclusion logic field combination, and generating a logic diagram structure with a connecting edge and a mutual exclusion mark. In the method, a threat path logic diagram is constructed by fusing field dependence, action timestamps and propagation hops, graph theory identification field mutual exclusion combination enhances cross-protocol attack chain analysis, and hidden Markov modeling state transition probability verifies time sequence continuity and path length. And performing dynamic time warping alignment on forward and reverse instruction sequences to extract semantic offset, overlapping rate and time sequence entropy, and performing non-linear score classification based on an isolated forest to detect an adversarial sample, topological structure analysis, time sequence verification, instruction alignment and non-linear classification to cooperatively identify a composite attack with field mutual exclusion and time sequence confusion.

The invention discloses a building material multi-parameter intelligent detection and analysis system, relates to the technical field of building material detection and monitoring, and is used for solving the problem of early abnormality recognition of a building material. According to the method, the dynamic change trend of the concrete material in the early maintenance stage is effectively described by constructing the multi-parameter state vector and combining the behavior evolution model, and the atypical abnormal behaviors such as early dehydration, structure segregation and reaction stagnation can be accurately recognized. Based on a non-linear classification and risk linkage mechanism, a response strategy matched with a behavior state is quickly generated, the intelligent level of detection and decision making is improved, a particularly introduced emergency triggering and short message quick response mechanism can quickly generate a minimum control instruction and directly reach a construction terminal after a high-risk state is identified, and the safety of the construction terminal is improved. Rapid intervention and action closed loop control are achieved, the recognition precision is higher, the response efficiency is higher, the anti-interference capacity is higher, and the method can be widely applied to concrete intelligent maintenance and structure early-stage quality control scenes.

The invention relates to the technical field of new energy vehicles, and discloses a new energy vehicle air conditioner compressor fault diagnosis system based on deep learning. The system comprises a data acquisition module, a model construction module, a model activation module and a result output module. The data acquisition module collects real-time operation data of the compressor to form an initial fault feature set containing a vibration spectrum feature sequence and a current fluctuation trend graph. The model construction module constructs a first diagnosis model and a second diagnosis model, the first diagnosis model reflects correlation mapping of vibration characteristics and mechanical faults, and the second diagnosis model comprises nonlinear classification rules of current characteristics and electrical faults. And the model activation module starts a corresponding target model and generates a diagnosis result according to the real-time vibration deviation and the current fluctuation abnormal data. And the result output module transmits the result to a vehicle control system to trigger a fault early warning signal or a maintenance suggestion instruction. The system can comprehensively cover mechanical and electrical fault diagnosis through dual-model collaborative operation, and is adaptive to a dynamic operation scene of the compressor.

The invention discloses a resume label generation method and device and a medium, and relates to the field of natural language processing, and the method comprises the steps: carrying out the preprocessing of an original resume text, and carrying out the word segmentation of original resume data into a plurality of text words; counting word frequencies corresponding to the text words, and endowing the text words with corresponding importance weights according to the word frequencies; vectorizing the text words to obtain corresponding word vectors, and aggregating the word vectors according to the importance weight to obtain resume text vectors corresponding to the original resume text; based on a preset hierarchical label library, performing multi-label classification on the text semantic vector through a nonlinear classification algorithm, and outputting a prediction label of the original resume text; and screening the label prediction values based on a preset threshold to obtain a final resume label set. By fusing word frequency weighted semantic representation and a nonlinear classification algorithm, the accuracy and adaptability of resume label generation are remarkably improved.

The invention provides a digital smart factory equipment operation management system based on deep learning, and the system comprises a data collection module, a data preprocessing module, a feature engineering module, a model training and fusion module, a dynamic optimization module, a fault judgment and early warning module and a data storage module which are in communication connection in sequence. According to the method, by providing a multi-algorithm fusion deep learning model architecture and integrating the advantages of time sequence prediction and a nonlinear classification model, the problem that complex data association cannot be mined by a single algorithm is solved, a precise data processing scheme adaptive to industrial data characteristics is designed, and the data quality and the characteristic discrimination degree are improved; and by constructing a model dynamic optimization mechanism, the problem of poor model generalization ability is solved, early-stage accurate early warning of faults is realized, the early warning advance time is prolonged, the non-planned shutdown probability is reduced, the real-time performance and robustness of the system are improved, and the method is adaptive to high-noise industrial scenes.

The invention discloses a fruit image classification method based on a wide area network dendritic neuron model, and the method comprises the steps: achieving the multi-scale feature extraction through parallel 1 * 1, 3 * 3 and 5 * 5 convolution and pooling branches, carrying out the fusion convolution, LeCannTanh activation and batch normalization, generating a feature representation through the combination of global average pooling and a lightweight feedforward network, and carrying out the classification of fruit images through the feature representation. And self-adaptive nonlinear classification is realized by means of synaptic-dendritic-somatic cell-axon hierarchical processing of dynamic dendritic neurons. Experimental results show that the method has the advantages of high precision, high robustness and low time delay in scenes with complex backgrounds, large intra-class differences and small samples, and is suitable for industrial applications such as intelligent sorting and automatic quality inspection.

The application discloses a kind of based on electrocardiogram and cardiac magnetic multi-scale fusion cardiac electrophysiological signal classification method and system, comprising: obtaining electrocardiogram signal and cardiac magnetic signal, and the original ECG, MCG signal is respectively filtered and denoised pretreatment, and through R wave positioning division heartbeat, after Fourier transform, obtain two-dimensional spectrogram, after feature extraction, respectively establish multi-domain cardiac function signal dataset;Convolutional recurrent neural network is constructed, and the synchronous processing of two different forms of cardiac electrophysiological signals is realized by double input model, shallow signal feature extracted by convolutional neural network is input into the time sequence feature extraction module formed by recurrent neural network, fusion ECG, MCG signal local saliency feature and time sequence feature, form the final ECG, MCG signal feature used for classification.Combined with cross-entropy loss function, the feature after fusion is input into nonlinear classification layer, using Softmax carries out model training, and outputs classification result.

An oral care device includes a head for caring for a user's oral cavity, the oral cavity including a plurality of oral regions, and an inertial measurement unit (IMU) operable to output signals that depend on the position and / or movement of the head. The oral care device includes a controller configured to receive from the IMU signals indicative of the position and / or movement of the head relative to the oral cavity, and to process the received signals using a trained non-linear classification algorithm to obtain classification data. The classification algorithm is trained to identify the oral region in which the head of the oral care device is located from the plurality of oral regions. The controller is configured to use the obtained classification data to control the oral care device to perform an action.

This invention discloses a method, device, and storage medium for bird species identification based on bird calls. The method includes: (1) acquiring several segments of bird calls and preprocessing them; (2) filtering the power spectrum of the bird calls using two different filter banks, extracting the coefficients of the filtered signals, and then combining the two sets of coefficients, the short-time energy of the bird calls, and the short-time zero-crossing rate of the bird calls to form the feature vector of the current bird calls; (3) constructing a nonlinear classification model and using a prey optimization method to find the optimal kernel function in the nonlinear classification model; (4) inputting the extracted bird call feature vector into the nonlinear classification model for learning; and (5) extracting the feature vector of the bird call to be identified and inputting the feature vector into the learned nonlinear classification model to identify the bird species. This invention has low complexity and high accuracy.

A method includes receiving, by a computing device, training data to train a neural network, wherein the training data comprises a plurality of inputs and a plurality of corresponding labels. The method also includes mapping, by a representation learner of the neural network, the plurality of inputs to a plurality of feature vectors. The method additionally includes training a kernelized classification layer of the neural network to perform nonlinear classification of an input feature vector into one of a plurality of classes, wherein the kernelized classification layer is based on a kernel which enables the nonlinear classification, and wherein the kernel is selected from a space of positive definite kernels based on application of a nonlinear softmax loss function to the plurality of feature vectors and the plurality of corresponding labels. The method further includes outputting a trained neural network comprising the representation learner and the trained kernelized classification layer.

The invention provides a mixed gas detection method and system based on a medical electronic nose, and the method comprises the steps: detecting whether a gas collection signal is received, and if yes, starting a DMA module of the medical electronic nose, and obtaining a gas detection value of a gas sensor. Preprocessing the gas detection value to obtain preprocessed data; and extracting a gas feature set of the preprocessed data, inputting the gas feature set into a multi-level decision framework for nonlinear classification to obtain a gas type, the gas type being used for representing the physical property of the user. According to the method, a decision tree is constructed based on the association number and association degree of various gas attributes and discrimination standard sets. And inputting a preliminary classification result into the improved random forest for secondary classification to obtain gas types, so that gas attributes with relatively low relevancy with the user physique are removed, gas attributes with relatively high relevancy with the user physique are highlighted, and the physique of the user is accurately classified in combination with various gases exhaled by the user.

The invention discloses a CNN and KAN fused electroencephalogram signal classification method, device and equipment and a storage medium, and relates to the technical field of artificial intelligence and neural networks, the method comprises the following steps: constructing a feature extraction encoder based on a first convolution block, a second convolution block and a distribution adaptation layer, and constructing a nonlinear classification decoder by using a KAN layer, the two are cascaded to form an initial electroencephalogram classification model; constructing a mixed data set containing source domain and target domain data, and training the initial model by combining a two-stage transfer learning strategy, a spline coefficient updating strategy and a mixed dynamic grid updating strategy to obtain a target electroencephalogram classification model; and performing data enhancement on to-be-classified multi-channel electroencephalogram data, inputting the to-be-classified multi-channel electroencephalogram data into the target electroencephalogram classification model, completing feature extraction, distribution alignment and nonlinear mapping, and outputting a classification result. According to the method, high-precision and stable electroencephalogram signal classification can be realized under the conditions of high noise of electroencephalogram signals, obvious individual difference and small samples.

The invention discloses a multi-sensor information fusion estimation method for state-dependent observation loss, and the method comprises the steps: constructing a system model which is a linear discrete dynamic system model and is used for representing a dynamically changing system state; training an observation loss discriminator, processing a complex observation rejection domain by the observation loss discriminator through a support vector machine, and converting a nonlinear classification problem into a linear convex optimization problem through a kernel function so as to discriminate whether a target state is in the observation rejection domain; and constructing a recursive fusion estimator based on an EM algorithm. According to the invention, the recursive estimator has high robustness and adaptability, can effectively deal with the complex situation of state dependence observation loss in various actual scenes, and significantly improves the precision of system state estimation.

The invention provides a mine ecological damage intelligent identification method and system based on remote sensing and a support vector machine, and relates to the technical field of mine ecological environment monitoring. The method comprises the steps of collecting multi-source remote sensing data to construct an initial data set, extracting comprehensive ecological damage features through a three-dimensional residual convolutional neural network, and realizing nonlinear classification by using a mixed kernel function support vector machine; meanwhile, an ecological damage index is calculated, a distribution diagram is generated, damage grades are divided by combining a dynamic threshold value, and a three-level early warning mechanism is triggered to generate restoration suggestions. According to the method, by introducing a deep learning model and a multi-source remote sensing data fusion technology, the expression ability of ecological characteristics is enhanced; through a support vector machine algorithm and kernel function mapping, the adaptability of the classification model in a complex scene is improved; through a dynamic threshold value and a multi-stage early warning mechanism, accurate grading and timely response of ecological damage are realized, and a scientific basis and technical support are provided for monitoring and restoration of a mine ecological environment.

The application relates to the technical field of computer vision and artificial intelligence, and discloses an intelligent image recognition and classification system based on deep learning, which comprises an image preprocessing module, a feature extraction module, a semantic modeling module, a classification decision module and a self-adaptive optimization module. The system firstly performs standardization processing on an original image, extracts multi-scale texture and edge features, combines an attention mechanism to model context semantic information, and forms high-dimensional representation; then multi-label output is realized through nonlinear classification mapping, and parameter weights are self-adaptively optimized according to a prediction error. The system has high robustness and generalization ability in a multi-class image scene, and is suitable for intelligent recognition and fine classification tasks in a complex image environment.

The application provides a mine ecological damage intelligent identification method and system based on remote sensing and support vector machines, relating to the technical field of mine ecological environment monitoring. The method includes collecting multi-source remote sensing data to construct an initial data set, extracting comprehensive ecological damage features through a three-dimensional residual convolutional neural network, and realizing nonlinear classification using a hybrid kernel function support vector machine. At the same time, the ecological damage index is calculated and a distribution map is generated, combined with a dynamic threshold to divide the damage level and trigger a three-level early warning mechanism to generate repair suggestions. The method enhances the expression ability of ecological features by introducing a deep learning model and multi-source remote sensing data fusion technology. Through the support vector machine algorithm and kernel function mapping, the adaptability of the classification model in complex scenarios is improved. Through the dynamic threshold and multi-level early warning mechanism, accurate grading and timely response to ecological damage are realized, providing a scientific basis and technical support for mine ecological environment monitoring and repair.

This invention discloses a method and system for analyzing the navigation behavior of ships in waterways based on big data, relating to the field of ship technology. This invention collects dynamic ship data and static waterway data, constructs an environmental semantic grid, and maps the dynamic data to generate semantic trajectory sequences. Based on the sequences, it calculates the basic spatiotemporal correlation value, extracts trajectory direction entropy using a local minimum spanning tree, and constructs a ship behavior feature vector. Based on physical limits, it constructs a dynamic pseudo-label dataset and uses an evolutionary algorithm based on a weighted ROC convex hull guidance strategy to iteratively optimize the parameters of the nonlinear classification decision function. Using the optimal parameter set, it constructs a decision function to identify abnormal ship behavior, calculates risk potential energy, and generates chain reaction warnings based on the risk transmission coefficient. This invention effectively integrates environmental semantics and entropy features, solving the problems of scarce abnormal samples and complex nonlinear feature identification without manual annotation, and achieving proactive and precise prevention and control of waterway collision risks.

The invention discloses a water transportation channel ship navigation behavior analysis method and system based on big data, and relates to the technical field of ships. Ship dynamic data and channel static data are collected, an environment semantic grid is constructed, and the dynamic data are mapped to generate a semantic trajectory sequence; calculating a basic space-time correlation value based on the sequence, extracting a trajectory direction entropy by using a local minimum spanning tree, and constructing a ship behavior feature vector; constructing a dynamic pseudo-label data set based on physical limits, and performing iterative optimization on parameters of the nonlinear classification judgment function by adopting an evolutionary algorithm based on a weighted ROC convex hull guide strategy; according to the method, environmental semantics and entropy value features are effectively fused, the problems of abnormal sample scarcity and complex nonlinear feature recognition under the condition that manual labeling is not needed are solved, and the method is suitable for ship abnormal behavior recognition. And active and accurate prevention and control of the water area collision risk are realized.

The invention discloses a photovoltaic power generation system operation state evaluation method and system, and the method comprises the steps: carrying out the subjective empowerment of an evaluation index of a photovoltaic power generation system to be evaluated through employing a subjective empowerment method, carrying out the objective empowerment of the evaluation index through employing a plurality of objective empowerment methods, carrying out the combined empowerment of a subjective weight and an objective weight, and obtaining a combined weight, the evaluation indexes and the combination weights form feature vectors, the feature vectors are input into a trained support vector machine model for operation state evaluation, an operation state evaluation result is obtained, the subjective and objective combination weighting method effectively fuses subjective experience and objective data driven weight information, the relative importance of all operation parameters is accurately quantified, and the operation state evaluation accuracy is improved. The support vector machine model has remarkable advantages when processing small samples and high-dimensional data by virtue of the strong nonlinear classification and generalization capability, the operation state can be efficiently and accurately judged according to multi-parameter characteristics, and the operation state of the photovoltaic power generation system can be efficiently and accurately evaluated by combining subjective and objective combination weighting and the support vector machine model.

The present application relates to a kind of unstructured data quality optimization method and system based on mathematical model, including receiving digital asset unstructured data, obtain data source diversity and inflow rate number, construct data quality related feature set;By feature extraction to feature set, the dynamic weight of unstructured data and nonlinear classification are calculated, data quality score and grade are obtained;Based on quality score and grade combination LSTM model carries out time-varying optimization, generates data quality optimization strategy.The present application realizes the dynamic evaluation and optimization of unstructured data quality by mathematical model, significantly improves digital asset management efficiency.Integrated data source diversity and real-time, accurately construct feature set;Dynamic weight and nonlinear classification improve the evaluation accuracy;Time-varying optimization and resource allocation priority design improve resource utilization.

The application provides a lung EGFR mutation recognition model based on CT sequence images, and belongs to the field of medical image analysis; the model comprises: a three-window preprocessing module configured to generate three-window CT sequence images according to original chest CT images, wherein the three windows comprise a lung window, a mediastinal window and a bone window; a sequence generation module configured to obtain three-window sequence images with consistent structures according to the three-window CT sequence images; a time sequence modeling module configured to perform frame-by-frame feature coding on the three-window sequence images, model the cross-slice spatio-temporal relationship of the coded slice features, and output sequence-level time sequence features fused with global spatio-temporal information; a sequence aggregation module configured to aggregate the sequence-level time sequence feature vectors, evaluate and weight the contribution of each slice to the final diagnosis, and generate patient-level fusion features; and a nonlinear classification module configured to output a prediction result of EGFR gene mutation according to the patient-level fusion features.

The application discloses an electroencephalogram signal classification method and device fusing CNN and KAN, equipment and a storage medium, relates to the technical field of artificial intelligence and neural networks, and comprises the following steps: constructing a feature extraction encoder based on a first convolution block, a second convolution block and a distribution adaptation layer, and constructing a nonlinear classification decoder by using a KAN layer, so that the initial electroencephalogram classification model is formed by cascading the two; a mixed data set containing source domain data and target domain data is constructed, and the initial model is trained by combining a two-stage transfer learning strategy, a spline coefficient updating strategy and a hybrid dynamic grid updating strategy, so that the target electroencephalogram classification model is obtained; the multi-channel electroencephalogram data to be classified is subjected to data enhancement, and then is input into the target electroencephalogram classification model, so that feature extraction, distribution alignment and nonlinear mapping are completed, and a classification result is output. The application can realize high-precision and stable electroencephalogram signal classification under the condition of high noise of electroencephalogram signals, significant individual differences and small samples.

The invention provides a lung EGFR mutation recognition model based on CT sequence images, and belongs to the field of medical image analysis. The model comprises: a three-window preprocessing module configured to generate a three-window CT sequence image according to an original chest CT image, the three windows comprising a lung window, a longitudinal partition window and a bone window; the sequence generation module is configured to obtain a three-window-level sequence with a consistent structure according to the three-window-level CT sequence image; the time sequence modeling module is configured to perform frame-by-frame feature coding on the three-window-level sequence, perform cross-slice space-time relationship modeling on the coded slice features, and output sequence-level time sequence features fused with global space-time information; the sequence aggregation module is configured to aggregate the sequence-level time sequence feature vectors and generate patient-level fusion features by evaluating and weighting the contribution degree of each slice to final diagnosis; and the nonlinear classification module is configured to output a prediction result of EGFR gene mutation according to the patient-level fusion features.

The invention discloses a rice cadmium pollution prediction method and system based on a multi-layer perceptron, and belongs to the field of cadmium pollution monitoring, and the method comprises the steps: obtaining rice soil sample data of various background regions, and marking positive and negative samples; the rice soil sample data is composed of multiple soil characteristics including the cadmium content; preprocessing the rice soil sample data; a multi-layer perceptron classifier is constructed and trained based on the preprocessed rice soil sample data, and the multi-layer perceptron classifier takes multiple soil features as input and takes whether rice cadmium exceeds the standard or not as an output target; the trained multi-layer perceptron classifier is used for conducting cadmium pollution prediction on rice soil data of the area to be predicted, and when the probability that rice cadmium exceeds the standard is predicted to be larger than a preset threshold value, it is indicated that rice cadmium has the standard exceeding risk; otherwise, the overproof risk of rice cadmium is low. According to the method, more comprehensive soil characteristics and an advanced MLP nonlinear classification model are introduced, so that the recognition capability of the rice cadmium exceeding standard sample is remarkably improved.

The invention relates to a data acquisition method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the steps of obtaining a description text of to-be-queried data; performing semantic coding on the description text according to the label mapping large model to obtain context semantic representation; according to a multi-hidden-layer pooling structure and a non-linear classification head of the label mapping large model, performing fusion processing and non-linear classification processing on the context semantic representation to obtain probability distribution corresponding to each label in a preset label set; and determining a target label based on the probability distribution, and obtaining target data corresponding to the to-be-queried data according to the target label. By adopting the method, the accuracy of obtaining the target data can be improved.

The application discloses a kind of wild animal habitat suitability evaluation method based on geographical weighted random forest classification model, belong to ecological information science, geographic information science, wild animal protection and machine learning technical field.The present application is aimed at the defects that existing global model cannot capture the spatial non-stationarity of environment-species relationship and traditional geographical weighted model is difficult to deal with nonlinear classification problem, constructs a spatial self-adaptive local random forest classification model system, completes the whole process of data library construction, spatial weight definition, optimal bandwidth optimization, global suitability prediction, result grading and driving factor analysis.The application significantly improves the habitat assessment accuracy in large-scale, high-heterogeneous habitat areas, can simultaneously analyze the dominant driving factors of species distribution in different regions, provides accurate scientific basis for wild animal habitat protection and nature reserve planning, and adapts to the evaluation needs of multiple types of wild animals.

The invention discloses an urban functional area surveying and mapping method based on multi-feature ensemble learning, and relates to the technical field of urban planning, remote sensing image processing and geographic information systems, and the method comprises the following steps: constructing a multi-modal double-flow adaptive fusion Transformer model MDSA-Former; defining a basic space unit; constructing a TSE feature set; the TSE feature set is input into an MDSA-Former, and a fused TSE'feature set is generated; an integrated learning classifier AdaBoost is adopted to classify the urban functional area units; and generating an urban functional area map to complete urban functional area surveying and mapping. According to the urban functional area surveying and mapping method based on multi-feature ensemble learning, self-adaptive fusion and nonlinear classification of multi-source heterogeneous features are efficiently achieved, and a high-precision and high-robustness technical scheme is provided for large-range and multi-category urban functional area automatic surveying and mapping.

The invention relates to the technical field of computer vision and artificial intelligence, and discloses an intelligent image recognition and classification system based on deep learning, which comprises an image preprocessing module, a feature extraction module, a semantic modeling module, a classification decision module and a self-adaptive optimization module. The system firstly performs standardization processing on an original image, extracts multi-scale texture and edge features, and models context semantic information in combination with an attention mechanism to form high-dimensional representation; and then multi-label output is realized through nonlinear classification mapping, and parameter weights are adaptively optimized according to prediction errors. The system has high robustness and generalization ability in a multi-category image scene, and is suitable for intelligent identification and refined classification tasks in a complex image environment.

The application discloses a mixture material dielectric property nonlinear mapping sorting method of high-dimensional manifold learning, and belongs to the technical field of intelligent sorting of regenerated plastics. Real-time scanning is performed on mixed plastic particles on a conveying channel to obtain real part and imaginary part data of complex dielectric constant at multiple characteristic frequencies, and a high-dimensional original characteristic vector is constructed. After white pre-processing, a local linear embedding algorithm is used to nonlinearly map the high-dimensional characteristics to a low-dimensional manifold space, and the essential dielectric characteristics of the material are retained. In the low-dimensional space, a nonlinear classification hyperplane is constructed based on a support vector machine to realize accurate identification of ABS plastics and impurity plastics. According to the identification result, the electrode voltage amplitude and polarity of a high-voltage electrostatic sorting machine are dynamically adjusted, a non-uniform electric field is constructed, and the motion trajectory of the particles is separated by using the coupling action of differential electrostatic force and gravity to make the target material and impurities fall into corresponding collecting hoppers. The application effectively distinguishes mixed plastics with similar dielectric properties, and improves the sorting precision and working condition adaptability.

The present application belongs to the field of supercharged boiler fault diagnosis, and relates to a supercharged boiler fault diagnosis method based on small sample learning and a training method and a testing method of a fault diagnosis model, aiming to solve the problem of data scarcity in supercharged boiler fault diagnosis, and the points include obtaining fault samples; dividing the fault samples into a training set and a validation set; randomly drawing fault samples from the training set to construct a batch of positive and negative sample pairs with the same number; inputting the positive and negative sample pairs into the twin deep network for training in batches; constructing a validation set and a validation support set; traversing the fault samples in the validation set; obtaining the fault diagnosis accuracy of the model, and the effect is to realize the supercharged boiler fault diagnosis with data scarcity belonging to small samples and nonlinear classification.